Skip to content
Other

Tev1 Bounty Hunter Qwen 0.8B

триаж BOLA/IDOR на edge-железе

2 min readupdated

Tev1 Bounty Hunter V2 — компактная модель на базе Qwen/Qwen3.5-0.8B-Base от zeroday01 для поддержки решений при проверке гипотез об уязвимостях веб-API. Она принимает HTTP-телеметрию и контекст гипотезы (BOLA/IDOR, BFLA) и возвращает структурированный JSON с оценкой, доказательствами и рекомендацией следующего шага. Рассчитана на ноутбуки с 4 ГБ VRAM и даже CPU.

Ключевые особенности

  • 100% BOLA/IDOR recall на held-out бенчмарке — 12/12; BFLA recall также 100%.
  • 0% false positives на защищённых endpoint'ах: корректно распознаёт 403 Forbidden.
  • ~832 МБ VRAM в 4-bit NF4; заявлен запуск на потребительском железе с 4 ГБ VRAM и на CPU.
  • Строгая JSON-схема с assessment, evidence, next step, abstention и uncertainty.
  • Без duplicate loops и invalid JSON: 0/60 в тесте, вывод — один top-level JSON-объект.
  • Advisory only: нет сетевых сокетов и автоматического выполнения атак; нужен human-in-the-loop.

Пример

import torch
from transformers import AutoProcessor, AutoModelForCausalLM

MODEL_ID = "zeroday01/tev1_cybersecurity"
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype=torch.bfloat16, device_map="auto")
# HTTP-телеметрию и гипотезу передайте в prompt
ПодробнееБенчмарк: BOLA/IDOR recall 100% (12/12), BFLA recall 100%, false-positive rate 0% на защищённых endpoint'ах.
Строгая JSON-схема — 80% (48/60), invalid JSON — 0%, duplicate loops — 0%.
В 4-bit NF4 нужно ~832 МБ VRAM; заявлены 4 ГБ VRAM или CPU.
Только advisory-режим, без сетевых сокетов и автоматических атак; выводы проверяет человек. Язык — английский.
Discuss Telegram 1 139

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.4 GB · ready files2.0 GB
  • FP8 / INT8recommendedweights 0.7 GB1.3 GB
  • 4-bitweights 0.4 GB0.9 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # zeroday01/tev1_cybersecurity
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download zeroday01/tev1_cybersecurity --local-dir ~/models/tev1_cybersecurity
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to tev1_cybersecurity
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Qwen3.5-0.8B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 9All on Hugging Face
Merges 1All on Hugging Face

About the model

zeroday01/tev1_cybersecurity
Author
zeroday01
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

1391

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments